Hello, I’m Kiran.
I build ML systems that work in the real world, and I think carefully about why they work (or don’t). Currently at Infinitus Systems, I design and ship NLP systems that automate healthcare phone calls at scale.
My current interests center on interpretability, reasoning, alignment, human behavioral modeling, and NLP for specialized domains. A little bit in detail here: what interests me in 2026.
Previously at Salesforce, Oracle, and PayPal. More in about and my resume.
Featured Topics
We are all architects now. Congrats on the promotion.
→
Reflections on the shift in my work identity from engineering to design and architecture in the age of agents.
5,000 Lines of Trust
→
Reflections on the shift in my work identity from engineering to design and architecture in the age of agents.
What Interests Me in 2026
→
My research interests in interpretability, alignment, and NLP for specialized domains—and why I think about ML throug...
The Art of Shipping ML
→
Reflections on shipping ML products: evaluation-centric design, the mirage of gold data, cost surface optimization, a...
Perplexity: The Poetry of Uncertainty
→
A deep dive into perplexity as a measure of uncertainty—the math, the intuition, and what it reveals about language m...
The Red Queen Runs, The Red King Waits
→
Evolutionary lessons for AI: the Red Queen's endless adaptation race vs. the Red King's slow cooperative strategy—and...
The Dance of Endurance
→
On long-distance cycling, the meditative flow of pushing through 200km, and the battle against the mind that wants to...
Let the fires burn
→
On protecting mental space as a finite resource—and the wisdom of letting some fires burn to clear the path.
Unplugged
→
A warm San Francisco morning at Alta Plaza Park: dogs, cats, tech bros, and the lost art of doing nothing.
The quarterly review — Q6
→
The final quarter: completing my Master's thesis on POI recommendation, graduating, and reflecting on two years at UCSD.
The quarterly review — Q5
→
17 credits of Learning Algorithms, Compilers, and TA-ing CSE 101. Insanely stupid, but rewarding in retrospect.
The quarterly review — Q4
→
Starting my Master's thesis on metric embeddings for recommendation. TA-ing algorithms, attending seminars, and a blu...
Day 1, Week 0, Year 2
→
Starting year two of grad school: lessons on taking the first step, prioritizing ruthlessly, and not going AWOL.
The quarterly review — Q3
→
A difficult quarter: health issues, heavy TA workload, and learning to pick myself up. Then off to San Francisco for ...
The quarterly review — Q2
→
Neural Networks, Web Mining, and Convex Optimization—a brutal winter quarter of coffee, projects, and late-night brai...
The quarterly review — Q1
→
First quarter at UC San Diego: Algorithms, AI, and Database Systems—learning to manage time in a fast-paced quarterly...